There is actually a bug when we do a single/multiple namespace installation mode.
At the moment we use OPERATOR_SCOPE_NAMESPACE env variable in order to specify the namespaces where the operator will reconcile and watch RabbitmqClusters.
In order to allow an operator group to work properly the csv needs to consume metadata.annotations['olm.targetNamespaces'] and pass it to our OPERATOR_SCOPE_NAMESPACE env variable.
I have implemented this through ytt as we already use it throughout the code.
There is actually a bug when we do a single/multiple namespace installation mode.
At the moment we use OPERATOR_SCOPE_NAMESPACE env variable in order to specify the namespaces where the operator will reconcile and watch RabbitmqClusters.
In OLM this is implemented through an operator group (og): https://docs.openshift.com/container-platform/4.8/operators/understanding/olm/olm-understanding-operatorgroups.html
In order to allow an operator group to work properly the csv needs to consume metadata.annotations['olm.targetNamespaces'] and pass it to our OPERATOR_SCOPE_NAMESPACE env variable.
I have implemented this through ytt as we already use it throughout the code.